Safety Alignment on CoSApien Public Prosecutor
0.941SafetyLlama-3.1-8B-Inst
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama-3.1-8B-InstOptimization Method=Reinforce++2026.05 | 0.941 | 4.115 | 3.9 | |
| Llama-3.1-8B-InstOptimization Method=Reward Distillation2026.05 | 0.91 | 4.365 | 4.035 | |
| CSRMOptimization Method=Reward Distillation2026.05 | 0.91 | 4.57 | 4.21 | |
| CSRMOptimization Method=Reinforce++2026.05 | 0.85 | 4.893 | 4.168 | |
| LlamaGuard-3-8BOptimization Method=Reinforce++2026.05 | 0.838 | 4.528 | 3.843 | |
| LlamaGuard-3-8BOptimization Method=Reward Distillation2026.05 | 0.765 | 4.705 | 3.59 |